Recommendation method and device based on deep learning

Abstract

The invention discloses a recommendation method based on deep learning, and the method comprises the steps: obtaining a plurality of user portraits and a plurality of commodity attributes, and locking a target user according to the plurality of user portraits; extracting the explicit features of the target user and the explicit features of the multiple commodity attributes and then processing the explicit features to generate a recommendation list; learning the hidden features of the target user and the hidden features of the plurality of commodity attributes by using a deep learning model, and predicting the score of the target user on the recommendation list according to the hidden features; pre-recommending the commodities in the recommendation list to the target user, determining whether the target user receives recommendation, and if yes, recommending related information of the commodities according to the score of the recommendation list and the priority. According to the method and device, multi-dimensional personalized screening is realized, the user autonomously selects whether to receive pushing or not, effective and accurate pushing is carried out, the dual requirements of the user and a merchant are met, and feature extraction based on deep learning has better noise immunity and effectiveness.

Description

technical field [0001] The present invention relates to the technical field of recommendation, in particular to a deep learning-based recommendation method and device. Background technique [0002] Classification of these models based on the form of input (methods with and without content information) and output (rating and ranking) has been proposed in the prior art. However, with the continuous emergence of new research results, this classification framework is no longer applicable, and a new inclusive framework is needed to better understand this research field. In the existing technology, the similarity between target users and approximate users is calculated. Degree, determine the approximate user with high similarity as the recommendation direction, and then recommend items to the target user by approximating the user's preferences. This is a recommendation method based on useless content information, and the recommendation method based on content information can be im...

AI Technical Summary

Problems solved by technology

However, there is no screening of recommended objects, and no consideration is given to whether users accept these recommendations. These are the reasons for the low recommendation efficiency, and they do not even consider the needs of users on the ground, and tap the potential needs of users. Recommendations with high accuracy and high satisfaction are often not achieved

Images